Androgen action in cell fate and communication during prostate development at single-cell resolution

The prostate develops from the endodermal urogenital sinus (UGS) that is derived from the caudal terminus of the hindgut called the cloaca (Cunha et al., 2018). Androgen signaling mediated through the androgen receptor (AR) is essential for prostate development (Gelmann, 2002). Mouse prostatic development initiates at embryonic day 17.5 (E17.5) from the UGS in response to rising levels of fetal testicular androgens (Cunha et al., 1987; Staack et al., 2003). Mutation of the Ar gene in testicular feminized (Tfm) mice results in the complete absence of prostate development (Cunha and Chung, 1981). During embryogenesis, the AR is initially detected in the urogenital sinus mesenchyme (UGM) before the initiation of prostate budding and morphogenesis. Subsequently, its expression extends to the urogenital sinus epithelium (UGE) after the initiation of prostatic budding and branching morphogenesis (Cunha et al., 1987; Cooke et al., 1991; Takeda and Chang, 1991). Tissue recombination studies further demonstrated that mesenchymal, rather than epithelial, AR signaling plays a decisive role in inducing development of the prostatic epithelium through paracrine regulation (Cunha and Lung, 1978; Cunha, 1984). However, the cellular properties of AR-expressing cells and the mechanisms by which stromal androgen signaling initiates and regulates other pathways and regulators, through mesenchymal-epithelial interactions during early prostatic development and morphogenesis, remain unclear.

During early development, budding and branching morphogenesis takes place in a variety of organs, including the prostate (Ochoa-Espinosa and Affolter, 2012; Iber and Menshykau, 2013; Varner and Nelson, 2014). Despite biological differences in cellular branch structure and function between organs and species, the developmental programs are largely conserved and are regulated by precise spatiotemporal epithelial-mesenchymal paracrine interactions (Thomson et al., 2002; Prins and Putz, 2008). Therefore, investigating mesenchymal androgen action in regulating prostatic budding and development will provide new and important insight into our understanding of cellular niches in branched organ development and their related human disorders.

The AR is a member of the nuclear hormone receptor superfamily (Chang et al., 1988). Androgen-induced AR transcriptional activity directly contributes to early prostate development and morphogenesis (Jenster et al., 1991). To trace AR-expressing cells and assess their functions during prostate development, we used gene-targeting approaches to generate a mouse Ar^IRES-Cre allele, which enables us to genetically mark AR-expressing cells and trace their fate and function in early prostate development. Using single-cell mRNA sequencing (scRNAseq) and other experimental approaches, we evaluated the cellular properties of AR-expressing cells at single-cell resolution, and characterized the signaling pathways and factors that contribute to early prostatic development and morphogenesis in mouse UGS tissues. Trajectory analysis further explores the identity of prostatic progenitors and the regulatory role of AR in prostatic differentiation. These data provide novel, high-resolution insight into cell fates and cell-cell interactions between prostatic mesenchyme and epithelium in early prostate development.

Using gene targeting approaches, we generated a new mouse model, Ar^IRES-Cre mice, by inserting Cre recombinase into the mouse Ar gene locus on the X chromosome through an engineering internal ribosome entry site (IRES) within the 3′ untranslated region (Fig. 1A). This advanced mouse tool allows for Ar promoter-directed expression of Cre recombinase while maintaining endogenous Ar expression. Both male and female Ar^IRES-Cre mice, including both heterozygotes and homozygotes, were viable and fertile, and presented no visible abnormalities. Female Ar^IRES-Cre/X mice were then crossed with male Rosa26^mTmG/+ (R26R^mTmG/+) reporter strains to generate R26R^mTmG/+:Ar^IRES-Cre/Y mice (Fig. 1B). The activity of Cre recombinase results in spontaneous recombination of the floxed reporter loci, generating a permanent genetic marker by switching from expression of membrane-bound tandem dimer Tomato (mT) to membrane-bound green fluorescent protein (mGFP) (Muzumdar et al., 2007) (Fig. 1B). Those genetically labeled cells will not only carry mGFP expression through their lifespan but will also pass it to their offspring, enabling us to trace the fate of AR-expressing cells and descendants.

Histological analysis of adjacent UGS sections prepared from male R26R^mTmG/+:Ar^IRES-Cre embryos between the ages of E12.5 and E17.5 showed normal morphology of the UGS at each time point (Fig. 1C1-C6). Co-staining of UGS tissues for vimentin, a mesenchymal cellular marker, and E-cadherin, an epithelial marker, visually separated epithelial and mesenchymal compartments in these sections (Fig. 1D1-D6). Interestingly, Cre-induced mGFP expression first appears mainly within the UGE in E12.5 UGS tissues (Fig. 1E1). The expression of mGFP continues to increase in the UGE and, starting at E13.5, is gradually revealed in the adjacent UGM (Fig. 1E2). The observation of mGFP expression in the UGE prior to that in the UGM is unexpected and suggests that Ar transcription actually occurs first in UGE cells, differing from the results reported previously (Cunha et al., 1987; Cooke et al., 1991; Takeda and Chang, 1991). Although the UGE appears almost completely mGFP positive by E14.5 and beyond, cells located in the protruding Wolffian duct remain mT positive through E17.5 (Fig. 1E5-G6, blue arrows). Immunofluorescent staining for mGFP showed a similar expression pattern to that observed in mTmG assays at different time points (Fig. 1F1-F6). Relatively weak AR staining appears within the UGM and beyond the UGE areas starting at E12.5 (Fig. 1G1,I1). It gradually becomes stronger (Fig. 1G2,G3) and extends to UGE areas, clearly visualized at E14.5 in both the UGE and UGM (Fig. 1G4-G6). At E12.5 and E13.5, cytoplasmic staining of AR appeared in cells with mGFP and DAPI (Fig. 1I1,I2,J1,J2, white arrows). However, uniform nuclear staining of AR clearly overlaps with DAPI in mGFP-positive UGE and UGM cells of E14.5 and elder UGS samples (Fig. 1I3-I6,J3-J6). The above results showed that Cre-induced mGFP expression mimics the expression pattern of endogenous AR but appears stronger than AR expression in UGS tissues. Additionally, the observation of mGFP prior to AR expression in UGE areas at E12.5 suggests that different translation processes through the 5′ cap and the IRES site take place for AR and Cre protein synthesis, respectively (Komar and Hatzoglou, 2011). Nevertheless, the current R26R^mTmG/+:Ar^IRES-Cre model provides a new tool for detecting and tracing AR-expressing cells.

To assess the characteristics of AR-expressing cells in early prostate development, we performed single-cell RNA sequencing (scRNAseq) analyses using UGS tissues isolated from male R26R^mTmG/+:Ar^IRES-Cre embryos (Fig. 2A). At E17.5, UGE and immediately surrounding mesenchyme reveal positive staining for Ar-Cre-induced mGFP expression, whereas much of the more distant stroma is mT positive, especially closer to the bladder (Fig. 2B). In addition, the Wolffian duct epithelial cells and their surrounding stromal cells, remain mT positive (blue arrow, Fig. 2B). Embryonic UGS tissues were dissected, and non UGS tissues, including the Wolffian duct, bladder and urethra, were removed (Fig. 2C); ∼11,000 viable cells were collected and used for sequencing with 10× Genomics Chromium Single Cell 3′ Solution (Fig. 2A). Following sequencing and alignment to the mm10 reference genome with an added mGFP sequence (Chen et al., 2015), 10,478 cells were included after the initial quality control in the analysis using the Seurat package (Butler et al., 2018). Cells then underwent filtering to remove potential empty droplets, doublets and low-quality cells with high percentages of mitochondrial RNA. After this final filtering step, 9440 cells with an average of 4976 genes and 31,214 UMI counts per cell were used for future analyses (Fig. S1A,B). When visualized in two dimensions based on unbiased transcription profiling using t-Distributed Stochastic Neighbor Embedding (tSNE), 21 distinct cellular clusters were identified (Fig. 2D). The cellular properties of these clusters were assessed using mesenchymal and epithelial cellular markers, including smooth muscle actin (Acta2), vimentin (Vim) and epithelial cell adhesion molecule (Epcam) (Fig. 2E and Fig. S2). Based on their expression patterns, epithelial and mesenchymal cell populations were determined and grouped (Fig. 2F). Seven stromal and four epithelial cell types were further identified (Fig. 2G-I) based on their transcription profiles in comparison with previously reported urogenital cellular markers and scRNAseq datasets from mouse and human prostate tissues (Georgas et al., 2015; Cunha et al., 2018; Henry et al., 2018; Kwon et al., 2019). Five highly expressed genes representing specific cellular properties in each cluster were shown (Fig. 2I). Based on these cellular markers, four epithelial clusters were identified as urogenital sinus epithelial (UGE), bladder epithelial (BLE), Wolffian duct epithelial (WDE), and other epithelial (OE) cells. Most of the epithelial cells were in the UGE cluster with high expression of Krt15, Krt4 and Tspan1. These UGE cells also showed higher expression of Ar and mGFP compared with other epithelial cells (Fig. S2). The cells in the BLE cluster displayed high expression of Upk3a, Upk2 and Sptssb, typical bladder cellular markers (Georgas et al., 2015; Habuka et al., 2015) (Fig. 2I), and low expression of Ar and mGFP compared with the UGE. The expression of Pax8, Pax2, and Hoxb7, WDE cell markers, was present in the WDE cluster (Narlis et al., 2007; Georgas et al., 2015) (Fig. 2B, Fig. S2A). Observation of small BLE and WDE cell clusters in the above samples was expected as those tissues cannot be completely removed from the UGS during sample preparation. Last, the OE cell cluster contains about 60 cells with high expression of Lrrn4, Gpm6a and Muc16, suggesting they have mesothelial cell properties (Kanamori-Katayama et al., 2011).

Seven stromal cell types were identified in the analysis. Most of the stromal cells possess fibroblast properties with Dcn, Fbln1 and Gsn expression (Guerrero-Juarez et al., 2019; Karpus et al., 2019). Smooth muscle cells are the second largest stromal cell population featuring high expression of Acta2, Tagln and Myh11 (Brun et al., 2015). Other minor stromal clusters were identified as leukocytes (Henry et al., 2018), endothelial cells (Fontijn et al., 2014), peripheral glial cells and neurons (Jessen and Mirsky, 2005; Jacob et al., 2014), and myoblasts (Ganassi et al., 2018), representing a similar cellular composition of the UGS at this timepoint, as reported previously (Georgas et al., 2015). Among those stromal clusters, Ar and mGFP are highly expressed in fibroblasts, with some limited expression in smooth muscle cells (Fig. 2I and Fig. S2A). These data provide a comprehensive landscape of the cell populations and properties of the male embryonic urogenital sinus at single-cell resolution.

Using the transcription profiles generated with scRNAseq, we further assessed the expression of endogenous Ar and Cre-induced mGFP at single-cell resolution in embryonic UGS tissues at E17.5. Most AR-expressing cells appear in the fibroblast and UGE cell clusters, and the signals from the former appear more intense than those from the latter (Fig. 3A-C). Cre-induced mGFP expression also appeared in both fibroblast and UGE cell clusters (Fig. 3B,C). The level of mGFP expression between the fibroblast and UGE cell clusters is comparable and is, in both cases, stronger than that of endogenous Ar (Fig. 3B,C). The different expression pattern between mGFP and endogenous Ar may be attributed to the fact that their transcription is regulated under different promoters, following the initial Cre-mediated activation of mGFP (Muzumdar et al., 2007). Analysis of cell distributions showed almost half of stromal and epithelial cells being mGFP and Ar positive (Fig. 3D). However, about one-quarter of stromal and one-third of epithelial cells appeared only mGFP positive, reflecting the transient nature of Ar expression in early prostate development. Few cells showed Ar expression without mGFP expression, suggesting Ar expression may have just initiated in those cells. Finally, about one-quarter of cells showed no detection of either Ar or mGFP expression in both the stroma and epithelia (Fig. 3D). Using immunofluorescent approaches, we validated the expression of endogenous AR and mGFP, as well as other cell markers in UGS tissues. AR and mGFP double-positive cells appear in both the UGE and UGM areas (Fig. 3F1-F4). Interestingly, staining for mGFP in the UGE cells appears stronger than UGM cells (Fig. 3F1). AR and mGFP double-positive cells also make up prostatic buds (arrows, Fig. 3F3,F4). The majority of UGE cells appear uniformly positive for E-cadherin and mGFP; however, weaker E-cadherin staining is present in areas along the exterior epithelial layer, especially the protruding prostatic buds (Fig. 3G1-G4). Co-staining of mGFP with vimentin was observed in the stromal cells directly surrounding the UGE cells (Fig. 3H1-H4). Smooth muscle cells appear to be separated from the UGE by several layers of fibroblasts and exhibit limited overlap with mGFP (Grishina et al., 2005) (Fig. 3I1-I4). The above co-immunofluorescence results provide in situ expression patterns in support of the previous scRNAseq transcription profile analyses.

The cellular properties of urogenital epithelial and mesenchymal cells were further evaluated using both scRNAseq and immunostaining. Krt8, a luminal epithelial marker appears to be expressed uniformly throughout epithelial cell clusters (Fig. 3E), supported by relatively uniform positive staining for CK8 within the UGE (Fig. 3J1,J2). Krt5, Krt14 and Trp63, prostatic basal cell markers, are localized to only the upper part of the UGE cluster in tSNE expression plots (Fig. 3E), and showed positive staining localized to the outer basal layer of the UGE and the prostatic buds (Fig. 3E,K1,K2,L1,L2,M1,M2). Staining of mGFP was also present in UGS areas and was overlaid with the above cellular markers. Interestingly, the expression of Fkbp5 and Serpinb2, two AR downstream target genes, showed different patterns. The expression of Fkbp5 appears in both mesenchymal and epithelial cells (Fig. 3E,N1,N2), whereas Serpinb2 is mainly expressed within the urogenital and bladder epithelial cells. Sonic hedgehog (Shh) is mainly present in the UGE and bladder cell clusters, whereas Gli1, a downstream target gene of Shh, appears in fibroblasts and smooth muscle cells (Fig. 3E), implicating their reciprocal regulation in early prostate development (Podlasek et al., 1999). Positive staining for Shh was also observed in the UGE cells (Fig. 3O1,O2). The above results from both tSNE expression and co-immunofluorescence staining provide new and consistent insight into the cellular properties of AR-expressing cells in embryonic UGS tissues.

To assess androgen signaling-initiated paracrine interactions between prostatic epithelial and mesenchymal cells, the epithelial cells were grouped and re-clustered, again yielding UGE, BLE, WDE and OE cell clusters (Fig. 4A and Fig. S3A-C). Given the biological significance of UGE cells in prostate development, we further re-clustered these cells. Re-clustering resulted in four clusters that were identified from the innermost to the outermost layer of the UGE, including superficial, intermediate, basal and budding UGE clusters based on urogenital epithelial cellular markers (Bhatia-Gaur et al., 1999; Grishina et al., 2005; Abler et al., 2011; Keil et al., 2012) (Fig. 4A,B and Fig. S3D,E). Although nearly all cells within these clusters are mGFP positive, levels of Ar expression vary, with the highest expression in the basal and intermediate UGE (Fig. 4C). Expression of prostatic budding markers, such as Nkx3.1, Wif1 and Bmp7, were mainly localized to the budding UGE cells, as well as Shh and Hoxd13, a transcription regulator involved in prostatic differentiation (Javed and Langley, 2014) (Fig. 4D). Interestingly, Etv4, which interacts with Shh and Fgf signaling in branching morphogenesis (Lu et al., 2009; Herriges et al., 2015; Zhang et al., 2016), was also highly expressed in the budding UGE cells. A full list of DEGs specific to the budding UGE was generated (Table S1). Immunohistochemistry analysis showed AR expression in both UGM and UGE areas, but weaker expression presented in budding epithelial cells than in adjacent stromal cells (Fig. 4E1,E2). Nkx3.1 and Etv4 expression appears selectively in the budding epithelial cells (Fig. 4E3,E4), whereas the expression of Bmp7, Hoxd13 and Wif1 was seen in both budding epithelial and surrounding stromal cells (Fig. 4E5-E7). Interestingly, staining for Shh appears noticeably weaker in the budding epithelial cells than in adjacent basal UGE cells (Fig. 4E8). A similar staining pattern has been observed in an earlier study (Podlasek et al., 1999). These data provide a clear breakdown of UGE cells at single-cell resolution during the onset of prostatic bud formation.

To explore the differentiation patterns of UGE cells, we performed trajectory analysis using Monocle's reverse graph embedding (Qiu et al., 2017). Three well-defined branches with budding UGE and superficial UGE cells concentrated at two opposing branch tips were generated with Monocle analysis. Basal and intermediate UGE cells made up most of the central region of the plot, displaying an overall arch from the innermost to the outermost UGE cell types (Fig. 5A,B). A mixture of basal and intermediate UGE cells formed the third, shorter branch tip (Fig. 5A). Prostatic basal cells, particularly, Zeb1-expressing basal cells, are thought to possess stem/progenitor cell properties and the ability to differentiate to prostatic luminal cells (Toivanen et al., 2016; Wang et al., 2020). Consistent with this recent study (Wang et al., 2020), the expression of Zeb1 appears to be concentrated at the end of the basal and intermediate UGE branch in the trajectory plot (Fig. 5C). Therefore, in the pseudotime plot, the point of origin was set as the small basal/intermediate branch displaying focal Zeb1 expression (Fig. 5B,C). This group of Zeb1 expressing cells also showed focal expression of other EMT related genes (Fig. 5C,D). A list of DEGs specific to this cell population was identified (Table S2). From that population, two distinct differentiation paths leading to prostatic bud cells and superficial UGE were revealed (Fig. 5B,C). These differentiation paths similarly start with a reduction in EMT markers, although the budding epithelial cells display a slight recovery of EMT-related genes, possibly contributing to their increased motility (Fig. 5D, right panel). Nkx3.1, Etv4 and Wif1 expression appears to be concentrated at the budding UGE branch tip (Fig. 5C,D). Although slightly diffuse, Hoxd13 and Bmp7 showed strongest expression in the budding UGE (Fig. 5C,D). These data suggest a regulatory role for Etv4, Wif1 and Nkx3.1 in differentiated budding cells while Hoxd13 and Bmp7 may act in initiating basal UGE cell differentiation to prostatic buds (Fig. 5C,D). Co-immunofluorescnece staining showed clear nuclear staining of ZEB1 in the majority of UGM cells; however, only varying cytoplasmic staining in UGE cells (Fig. 5E1,E5). Epithelial ZEB1 staining was also overlaid slightly with CK8 but not Trp63 staining in UGE cells (arrows, Fig. 5E4,E5,E7,E8). The above observations appear similar to previous reports showing cytoplasmic staining of ZEB1 in epithelial cells (Spaderna et al., 2006; Goscinski et al., 2015). Using SCENIC upstream regulator analysis (Aibar et al., 2017), we further assessed the driving factors contributing to UGE cell differentiation (Fig. S3F). The top 20 transcription factors (TFs) were identified in each UGE cell type, including Wnt (Sp5 and Tcf7) and Notch signaling-related regulators (Zfp64 and Hey1), as well as ETV family members (Fig. S3F). Taken together, these data determine the differentiation status of UGE cells and explore the cell properties of urogenital epithelial progenitor cells at single-cell resolution.

We similarly re-clustered the stromal cells to gain deeper insight into their cell characteristics (Fig. 6A). Eight cell types were identified (Fig. 6B). The newly identified cell type, pericytes, were previously grouped within the smooth muscle cells (Fig. 2G). High expression of Ar and mGFP appeared primarily in fibroblasts (Fig. S4C). As urogenital sinus fibroblasts are directly adjacent to the UGE and involved in regulating the epithelial-mesenchymal paracrine interactions during early prostate development (Grishina et al., 2005; Abler et al., 2011), we re-clustered them to further characterize their cellular properties. Of the nine resulting clusters, one showed high expression levels of cell proliferation-related genes (Fig. 6C). Using cell-cycle regression, we removed those genes from the principal component analysis (PCA). Following this regression, re-clustering produced eight fibroblast clusters (Fig. 6D,E) that were grouped into four subtypes, including progenitor-like (ProGf), Col14a1-expressing (Col14a1f), peri-epithelial (PeriEf) and other fibroblasts (OF) (Fig. 6F).

We first identified Bmp7-, Wif1- and Wnt5a-expressing fibroblasts following reports that they are directly surrounding prostatic buds and involved in early prostate development and morphogenesis (Grishina et al., 2005; Huang et al., 2009; Keil et al., 2012). The expression of Bmp7, Wif1 and Wnt5a appear mainly in clusters F4 and F5, and thus were identified as peri-epithelial fibroblasts (Fig. 6F,G), which also showed the highest level of Ar expression among the fibroblast clusters (Fig. 6F). In co-immunofluorescence analyses, Bmp7, Wif1 and Wnt5a proteins were also detected in the epithelial buds and surrounding stromal cells and overlaid with AR (Fig. 6I′-K′). In addition, as observed in the scRNAseq, staining for AR appeared strongest in the UGM adjacent to the UGE, whereas the lowest expression levels were observed in the prostatic buds (Fig. 6I′-K′). The expression of Bmp7, Wif1, and Wnt5a along with elevated AR appeared in peri-epithelial fibroblasts as well as in the budding UGE cells, implicating their participation in epithelial-mesenchymal paracrine interactions during prostatic epithelial budding and development.

We further assessed the relationships between different fibroblast clusters during early prostate development using trajectory analysis. Four major branch tips were observed in the trajectory plot generated using Monocle (Fig. 6H). Their differentiation states were predicted using CytoTrace (Fig. S4D-F). Based on this, a group of potential progenitor fibroblasts, clusters F1 and F2, were identified at one of the branch tips, and set as the starting point for pseudotime analysis (Fig. 6H). The peri-epithelial fibroblasts were grouped together at another branch tip (Fig. 6H). A third branch tip featured a group of fibroblasts, primarily from cluster F3, with high expression of Col14a1, Clec3b and Anpep (Fig. 6F,H), similar to Col14a1 expressing fibroblast populations reported in the lung and skin (Xie et al., 2018; Guerrero-Juarez et al., 2019). The identity of the fourth branch tip is unclear and possesses a mixture of the fibroblast subtypes. The expression of Bmp7 and Wif1 were mainly observed in the peri-epithelial fibroblast branch tip, while Wnt5a, Fgfr2 Hoxd13 and Gli1 expression appeared in all fibroblast branch tips but with focal expression on some individual tips (Fig. 6L). These trajectory analyses explored the cellular properties and differentiation status of prostatic fibroblasts.

To assess the regulatory role of the peri-epithelial fibroblasts in early prostate development, we examined the DEGs comparing the transcriptomes of peri-epithelial fibroblasts against the other fibroblasts (Fig. 7A and Table S3). Using gene set enrichment analysis (GSEA) with different databases, a number of enriched pathways, previously reported to directly associate with branching morphogenesis, were identified (Fig. 7B), including Wnt/β-catenin, Bmp, hedgehog, androgen and retinoic acid-related signaling pathways. The top 40 DEGs from the above analysis displayed elevated expression of peri-epithelial fibroblast markers Wif1, Bmp7 and Wnt5a, along with hedgehog downstream targets, Foxf1, Cxcl14 and Dner (Madison et al., 2009; Yu et al., 2009), and Wnt downstream targets, Lef1 and Tcf4 (Fig. 7A). Using IPA, a subset of upregulated AR downstream genes was identified from the above DEGs, suggesting a regulatory role for AR in peri-epithelial fibroblasts (Fig. 7C). Using chromatin immunoprecipitation-quantitative polymerase chain reaction (ChIP-QPCR), we demonstrated an enrichment in the regulatory regions of AR downstream target genes (Nash et al., 2019), including Fgfr2, Fgf7, Sgk1, Prkca and Ets2 in AR antibody immunoprecipitated DNA samples isolated from E17.5 UGM cells in comparison with input samples (Fig. 7D). The transcriptional regulators enhanced in peri-epithelial fibroblasts were further identified using SCENIC (Aibar et al., 2017) master regulator analysis (Fig. 7E), including seven Fox and Hox family TFs, five Wnt-related TFs, including Lef1 and Tcf4, as well as Sox9 and Sox4. In addition, many of these TFs have been identified as AR-associated proteins and co-regulators (http://androgendb.mcgill.ca/ARinteract.pdf; Fig. 7E). Given proximity of peri-epithelial fibroblasts to the budding UGE cells, we characterized the interactomes between these two clusters to explore the paracrine interactions between the two cell types using SingleCellSignalR methods (Cabello-Aguilar et al., 2020). Predicted ligand-receptor interactions related to Fgf, Tgfβ, Shh, Wnt and Notch signaling pathways appeared in both directions between peri-epithelial fibroblast and budding UGE cell clusters (Fig. 7F). These data demonstrate the important role of peri-epithelial fibroblasts in inducing prostate epithelial development and provide an atlas for further investigating the role of AR and other regulators in controlling prostate initiation and development.

Using mouse genetic tools, we further assessed the role of AR in UGM and UGE cells during early prostate development. As illustrated in Fig. 8A,B, selective deletion of Ar expression and activation of mGFP expression in Gli1-expressing cells in the UGM, and in Trp63- and Krt8-expressing cells in the UGE was achieved by administrating TM at E13.5 in R26R^mTmG/+:Ar^L/Y:Gli1^CreERT2/+, R26R^mTmG/+:Ar^L/Y:Trp63^CreERT/+ and R26R^mTmG/+:Ar^L/Y:Krt8^CreERT2/+ mice, respectively (Fig. 8B). Histological and immunohistochemical analyses showed normal prostate budding with positive staining of E-cadherin and Nkx3.1 in UGS tissues isolated from E18.5 wild-type mice (Fig. 8C1-C3). Immunofluorescence staining of mGFP in UGS tissues of R26R^mTmG/+:Gli1^CreERT2/+, R26R^mTmG/+:Trp63^CreERT/+ and R26R^mTmG/+:Krt8^CreERT2/+ mice showed Gli1-expressing cells in the UGM, and Trp63- and Krt8-expressing cells in different UGE compartments (Fig. 8D1-D3), demonstrating their mesenchymal or epithelial properties, respectively. Examining E18.5 UGS tissues of R26R^mTmG/+:Ar^L/Y:Gli1^CreERT2/+ mice revealed no prostatic bud formation (Fig. 8E1-E3′), consistent with previous reports (Le et al., 2020). In contrast, no significant defects appeared in UGS tissues isolated from Ar^L/Y:R26R^mTmG/+:Trp63^CreERT/+ and Ar^L/Y:R26R^mTmG/+:Krt8^CreERT2/+ embryos (Fig. 8H1,H1′,K1,K1′). Positive staining of both E-cadherin and Nkx3.1 also appear in the above UGS tissues (Fig. 8H2,H3′,K2,K3′). These data further demonstrate the significant role of mesenchymal AR signaling in prostate budding and development. Positive staining for Lef1, Tcf4, Hoxd13 and Sox9, which we demonstrated to be enhanced in peri-epithelial fibroblasts (Fig. 7A,C), was observed in the UGM or in both UGM and UGE cells of Ar^L/Y:R26R^mTmG/+:Trp63^CreERT/+ and Ar^L/Y:R26R^mTmG/+:Krt8^CreERT2/+ tissues (Fig. 8I1-I4,L1-L4). Both Fgf7 and Fgfr2 (with VIM co-staining) are also present in Ar^L/Y:R26R^mTmG/+:Trp63^CreERT/+ and Ar^L/Y:R26R^mTmG/+:Krt8^CreERT2/+ UGM cells (Fig. 8J1,J2,M1,M2). However, little to no staining for those proteins was detected in Ar^L/Y:R26R^mTmG/+:Gli1^CreERT2/+ UGS tissues (Fig. 8F1-F4,G1,G2). Taken together, these data demonstrate that selective deletion of AR in mesenchymal Gli1-expressing cells, but not in UGE cells, impairs prostatic budding and diminishes the expression of key transcriptional and paracrine regulators in embryonic UGS tissues. This also provides proof-of-principle evidence supporting the scRNAseq data in this study.

Currently, the molecular mechanisms underlying fetal androgen-induced interactions with other signaling pathways in early prostate development remain unclear. Specifically, the identity of prostatic AR-expressing cells that convey androgen signaling and control early prostate induction and development are unknown. In this study, using the novel mouse strain R26R^mTmG/+:Ar^IRES-Cre/Y, we assessed AR-expressing cell properties and investigated the molecular basis of AR-mediated signaling in initiating prostatic mesenchymal-epithelial interactions during early prostate development. Interestingly, Cre-induced mGFP expression through endogenous AR transcription appeared in the E12.5 UGE rather than the UGM. This observation provides the experimental evidence that AR transcription actually occurs first in the UGE although the expression of endogenous AR proteins is first detected in the UGM at E13.5 (Cooke et al., 1991). This observation is also consistent with an early in vivo hybridization study showing that AR transcription starts at E12.5 in UGE and UGM cells (Crocoll et al., 1998). Cre-induced mGFP expression appears to mimic the expression pattern of endogenous AR but in a much more sensitive manner, which may be due to the high efficiency of Cre-mediated recombination and the different translational processes between the 5′ cap and IRES site for synthesizing AR and Cre proteins (Komar and Hatzoglou, 2011). Therefore, the Ar^IRES-Cre model is a valuable tool for tracing AR-expressing cells when combined with reporter alleles, and for perturbation and manipulation of androgen signaling through the incorporation of other floxed alleles.

Using the Ar^IRES-Cre model with scRNAseq and other experimental approaches, we assessed the male E17.5 UGS epithelial and mesenchymal cell types. The expression of endogenous Ar and Cre-initiated mGFP was mainly observed in the UGE and fibroblast cell clusters, suggesting the biological significance of androgen signaling in these cells. We further separated and re-clustered these cell populations individually to visualize them at the highest resolution. Based on highly specific expression of prostatic budding markers, including Nkx3.1, Wif1 and Bmp7, we identified a cluster comprising budding epithelial cells. In addition, the expression of Etv4, a key regulator in budding/branching processes in lung and kidney branching morphogenesis (Lu et al., 2009; Costantini, 2010; Herriges et al., 2015) and in promoting epithelial-mesenchymal transition (EMT) and prostate cancer progression (Pellecchia et al., 2012; Aytes et al., 2013), was also identified as a novel marker of budding UGE cells. Using trajectory analysis with Monocle and CytoTrace software, we also explored the cellular identity of prostatic epithelial progenitors in E17.5 UGS tissues. A potential progenitor population comprising Zeb1-expressing cells within basal/intermediate UGE clusters was identified, showing opposing pathways of differentiation leading to either budding or superficial UGE clusters (Wang et al., 2020). These results provide new insight into the cellular properties of budding and UGE progenitor cell populations, and provided unbiased characterization of the transcriptomes of these unique epithelial cells, helping to identify the signaling pathways and regulators acting in prostatic epithelial differentiation and development.

We also characterized urogenital stroma to better understand mesenchymal AR signaling in regulating epithelial-mesenchymal paracrine interactions during prostate epithelial development. As urogenital sinus fibroblasts are directly adjacent to the UGE, we specifically focused on peri-epithelial fibroblasts, which were shown to have the highest AR expression and expression of Bmp7, Wif1 and Wnt5a, regulators of prostatic budding (Grishina et al., 2005; Huang et al., 2009; Abler et al., 2011). Upregulation of developmental signaling pathways, including Wnt/β-catenin, Bmp, retinoic acid, hedgehog and androgen signaling were identified in these fibroblasts based on the DEGs when compared with other fibroblasts. IPA analyses showed the upregulation of AR downstream target genes in these peri-epithelial fibroblasts. ChIP Q-PCR further demonstrated an enrichment of AR binding in the regulatory regions of AR target genes, reported in AR ChIPseq data from P0 rat ventral and dorsolateral prostate lobes (Nash et al., 2019), in AR-antibody immunoprecipitated E17.5 UGM samples. Analysis of the interactions between the prostatic bud epithelia and peri-epithelial fibroblasts, using SingleCellSignalR (Cabello-Aguilar et al., 2020), further identified paracrine interactions related to Wnt, Bmp, Notch and hedgehog signaling pathways. The above data provide the molecular basis for mesenchymal AR in regulating paracrine interactions between mesenchymal and epithelial cells at single-cell resolution during prostate early development.

Last, using mouse genetic tools, we further evaluated the role of epithelial and mesenchymal AR in prostate development. Consistent with previous reports (Le et al., 2020), impairment of prostatic budding appeared in E17.5 UGS tissues with Gli1-CreER-induced Ar deletion but no significant defects showed in counterparts with Krt8- or Trp63-CreER-induced AR deletion, further demonstrating the crucial role of mesenchymal AR in prostate development. Interestingly, Zeb1 expression was observed mainly in mesenchymal Gli1-expressing cells, although epithelial Zeb1-expressing cells with co-expression of Krt8 also showed weak expression of Gli1 in tSNE plots (Fig. S5A-F). These observations raise the possibility of either Gli1-expressing cells as the origin of Zeb1-expressing progenitors or Gli1 expression occurring in epithelial cells during embryonic prostate development. Therefore, more effort should be devoted to defining the cellular properties of Gli1- and Zeb1-expressing cells as prostatic stem cells and/or their niche cells in prostate development. Taken together, this study provides significant insight into the regulatory role of mesenchymal AR and provides a wealth of data for future studies of epithelial-mesenchymal interactions in prostatic development and disease.

All animal procedures were approved by the City of Hope Institutional Animal Care and Use Committee (IACUC). Rosa26^mTmG/+ (R26^mTmG/+) reporter mice were kindly provided by Dr Liqun Luo (Stanford University, CA, USA) (Muzumdar et al., 2007). Experimental mice were generated by intercrossing Ar^IRES-Cre/X female mice with R26^mTmG/+ male mice. Gli1^CreER mice were obtained from Jackson Laboratories (stocks 007913). Ar^Lox/Y mice were obtained from Dr Guido Verhoeven (De Gendt et al., 2004). Trp63^CreERT2 and Krt8^CreERT2 mice were generated as described previously (Lee et al., 2014; Zhang et al., 2012). To elicit genetic recombination, mice were intraperitoneally injected with 125 µg/g body weight of tamoxifen (TM, Sigma) suspended in corn oil (Sigma) as previously reported (Lee et al., 2015). Genotyping was performed using the primers provided (Table S4). The targeting vectors for generating Ar^IRES-Cre mice were designed based on a previous study for making Ar^{IRES-PLAP-IRES-nLacZ} mice (Shah et al., 2004). The IRES-Cre fragment was generated using PCR approaches from the IRES-mCherry plasmid (Addgene #80139) and assembled accordingly. The fragment was introduced to the region after the stop codon of the mouse Ar gene. Three guide RNAs (gRNA) were designed at the above insertion site, one gRNA (AAGTGCCCAAGATCCTTTCT) was chosen after in vitro cutting efficiency validation. The gRNA was synthesized using in vitro transcription approaches, and CAS9 protein was purchased from PNA Bio. The above IRES-cre donor DNA fragment (dDNA) was cloned into the targeting vector and bracketed with left and right homologous arms (∼1 kb each), which were generated by PCR using C57Bl6/J mouse genomic DNA as template with the following pairs of primers: AR-5F, TTCCAGTGGATGGGCTGAAAAATC; AR-5R, ACGCGTTCTTCACTGTGTGTGGAAATAGATGGGCTTGACTTTGCCAGAAAGGATCTTGGGCACTTG; AR-3F, TTTGGAAACCCTAATACCC; AR-3R, CAAAGAGTCAGACCTTTCC. A point mutation was added by the AR-5R primer at gRNA PAM to prevent gRNA cutting donor DNA. C57BL/6J mice (4-6 weeks old) were used for zygote collection. CD1 females were used for injected embryo recipients. Guide RNA, CAS9 protein and donor DNA were mixed before injection into mouse zygotes at a final concentration of 5 µg/µl gRNA, 15 µg/µl CAS9 protein and 3 ng/µl dDNA. Injected zygotes were implanted into CD1 foster mothers. The genomic DNA samples were isolated from the offspring and analyzed by genomic PCR and DNA sequencing. Confirmed Ar^IRES-Cre founder mice were then backcrossed with wild-type C57BL/6J for two to three generations to delete mosaicism resulting from Crispr/CAS9-based gene-editing approaches in mice.

Mouse tissues were fixed in 10% neutral-buffered formalin (American Master Tech Scientific) and processed in paraffin or processed to OCT following cryo-protection in 30% sucrose in 1× PBS (pH 7.3) at 4°C overnight as previously described (Lee et al., 2015). Following embedding in paraffin wax or OCT, tissue blocks were cut to 4 μm and 5 μm serial sections, respectively, and used for Hematoxylin and Eosin staining for further histological analysis (Lee et al., 2015). Immunohistochemistry was performed as previously described (Lee et al., 2015). In brief, tissue sections were rehydrated through a decreasing ethanol gradient. Heat-induced epitope retrieval was performed using a microwave to boil slides in 0.01 M citrate buffer (pH 6.0) followed by 15 min in 0.3% H₂O₂ in methanol. Next, tissue sections underwent blocking for 1 h at room temperature in 5% goat serum in 1× PBS (pH 7.3) followed by incubation with primary antibodies in 1% goat serum in PBS overnight at 4°C. Tissue sections were then washed with PBS and incubated with streptavidin ligated to horseradish peroxidase (Strep-HRP) (SA-5004, Vector Laboratories, 1:500 dilution) for 30 min and developed with 3,3′-diaminobenzidine (DAB) kit (SK-4100, Vector Laboratories). Tissue sections were counterstained in 5% Harris Hematoxylin and dehydrated through an increasing ethanol gradient. Coverslips were mounted using Permount Medium (SP15-500, Fisher Scientific).

Immunofluorescence staining was performed using similar procedures as described for immunohistochemistry; however, the use of H₂O₂ in methanol was excluded. Slides were developed with fluorescence-conjugated secondary antibodies and then mounted with coverslips using Vectashield Mounting Medium with DAPI (H-1200, Vector Laboratories). For detecting membrane-bound Tomato (mT) and membrane-bound green fluorescent protein (mGFP) signal, the staining was performed as previously described (Lee et al., 2015). Briefly, sections from the OCT-embedded tissues were washed with PBS (pH 7.30) and were developed with Vectashield Mounting Medium with DAPI (H-1200, Vector Laboratories). Antibodies used for both immunohistochemistry and immunofluorescence are listed in Table S5.

Images of Hematoxylin and Eosin and immunohistochemistry were acquired on an Axio Lab A1 microscope using 5×, 10×, 20× and 40× Zeiss A-Plan objectives with a Canon EOS 1000D camera and using Axiovision software (Carl Zeiss). Images of immunofluorescent staining and mTmG signals were acquired on a Nikon ECLIPSE E800 Epi-Fluorescence Microscope using 20× and 40× Nikon Plan Fluor objectives with a QImaging RETIGA EXi camera and using QCapture software (QImaging).

Female mice were placed with males overnight for mating, with the following day considered embryonic day (E) 0.5 if a vaginal plug was detected. Two individual sets of sc-RNA-seq experiments were performed using different littermates in this study. The urogenital sinuses of embryos were isolated at E17.5 and placed in DMEM with 10% FBS, 10 nM DHT, 5% Nuserum and 25 μg/ml insulin. The UGS was separated from the bladder at the bladder neck, and distal regions of the urethra and Wolffian ducts were removed (Fig. 2C). Once trimmed, UGS tissues were dissociated to single cells by digestion using 1 mg/mg collagenase at 37°C for 90 min and TrypLE (Gibco) at 37°C for 15 min. Approximately 11,000 viable cells obtained from male R26R^mTmG/+:Ar^IRES-Cre embryos (n=4) were used for sequencing. Library preparation was performed using 10× Genomics Chromium Single Cell 3′ Solution with v3 chemistry following the manufacturer's protocol (10x Genomics). The library purity and size was validated by capillary electrophoresis using 2100 Bioanalyzer (Agilent Technologies). The library quantity was measured fluorometrically using Qubit dsDNA HS Assay Kit from Invitrogen. The libraries were sequenced with a NovaSeq 6000 instrument (Illumina) to a depth of ∼130K reads per cell. Raw sequencing data were processed using the 10× Genomics' Cell Ranger pipeline (version 3.1.0) to generate FASTQ files and aligned to the mm10 genome with an added mGFP sequence (Zhang et al., 1996) to generate gene expression counts. Following alignment and initial quality control, a filtered feature bar coded matrix, including a total of 10,478 cells, was uploaded to R (3.6.1) using the Seurat package (3.1.3.9002) (Butler et al., 2018). Cells then underwent further filtering to remove potential empty droplets and doublets (1000<nFeature_RNA<9000) as well as low-quality cells with high percentages of mitochondrial RNA (percent.mt<15) (Fig. S1A,B). After this final filtering step, 9440 cells with an average of 4976 genes per cell and 31,214 UMI counts per cell were used for future analyses. Two individual sets of sc-RNA-seq experiments were performed using different littermates in this study.

For the visualization and analysis of scRNAseq results, normalized and scaled data were clustered using the top significant principal components (30) of 3000 highly variable genes with a resolution of 0.5 using Seurat (Fig. S1C,D). Trajectory analysis and generation of pseudotime was performed by converting Seurat objects into CellDataSet format. Following conversion of the data to a CellDataSet, trajectory analysis was performed using the Monocle2 package (2.12.0) (Qiu et al., 2017) in R. Identification of stem/progenitor populations was supported through the use of CytoTRACE (https://cytotrace.stanford.edu/) (Gulati et al., 2020). Interactome (ligand-receptor) analysis and visualization were performed using the SingleCellSignalR package (0.0.1.6) in R (Cabello-Aguilar et al., 2020). Pathway analysis was performed using Gene Set Enrichment Analysis (GSEA 4.0.3). Last, single-cell upstream regulator analysis was performed using Ingenuity Pathway Analysis (Qiagen, Version 51963813) and the SCENIC package (1.1.2.2) in R (Aibar et al., 2017).

The urogenital sinuses of embryos were isolated at E17.5. Mouse urogenital sinus mesenchyme (UGM) tissues were collected as reported previously (Xin et al., 2003). Briefly, the UGS was collected in 10 ml DMEM 10% FBS and 10 nM DHT, and rostral and caudal parts of UGS tissues and the Wolffian duct were removed. Trypsin was removed carefully, and digestion was stopped with 250 μl of DMEM, 20% FBS and 10 nM DHT followed by two washes with same amount of medium. The digested UGS was separated to UGE and UGM using a fine needle under a microscope. Mouse UGM (n=6) tissues were minced and incubated with 1% formaldehyde for 15 min and quenched with 0.150 M glycine for 10 min. Samples were washed sequentially with ice-cold PBS, and resuspended in cell lysis buffer [50 mM Tris-HCl (pH 8.0), 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% NP-40 and 0.25% Triton X-100], and then homogenized. The chromatin was sheared in nuclear lysis buffer [10 mM Tris-HCl (pH 8.0), 1 mM EDTA, 0.5 mM EGTA and 0.2% SDS] to an average size of 200-500 bp by sonication, diluted threefold in ChIP dilution buffer [0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-HCl (pH 8.1) and 167 mM NaCl], and then subjected to immunoprecipitation by magnetic protein G beads (Invitrogen) conjugated with AR (ab74272, Abcam). Crosslinks were reversed and chromatin DNA fragments were analyzed by real-time qPCR with specific primers (Table S6).

Differentially expressed gene lists were determined using a Wilcoxon Rank Sum test, with changes in expression at P<0.05 defined as significant. As recommended by the GSEA User Guide, pathways with FDR<0.25 were considered significant in exploratory GSEA pathway analysis. Source data underlying Figs 3D, 7B,C,D,F are provided in Tables S7-S11, respectively.